With Poland as a Counter-Factual
Albino Jos´
e Rocha Muchacho
Dissertation submitted as partial requirement for the conferral of: Master in Economics
Supervisor:
Prof. S´ergio Lagoa, Prof. Auxiliar, ISCTE-IUL Escola de Ciˆencias Sociais e Humanas, Departamento de Economia Pol´ıtica
Co-supervisor:
Prof. Ryszard Kokoszczy´nsk, Assistant Professor, University of Warsaw, Faculty of Economic Sciences, Division of Quantitative Finance
This dissertation aims to present and discuss the empirical evidence of the trans-mission mechanisms of monetary policy in Portugal before and after the Euro and use Poland as a counter-factual.
The analysis focus in the interest rate channel of transmission of monetary policy. The VAR model is used to simulate the effects of a change in interest rates in the economy.
Two conclusions can be drawn, firstly there is a clear difference in the Portuguese responses and explanatory power of interest rate before and after the adoption of Euro, since the responses are stronger in the Pre-Euro model. Finally, there are similarities between Portugal and Poland as they share similar responses.
Keywords: Exogenous Block, Monetary Transmission Mechanism, VAR JEL code: E53 and E58
Resumo
O objetivo desta disserta¸c˜ao ´e apresentar e discutir as evidencias emp´ıricas sobre os mecanismos de transmiss˜ao da pol´ıtica monet´aria em Portugal, antes e depois do Euro e comparar com a Polonia, que ´e usada como contra factual.
A an´alise foca-se no canal da taxa de juro do mecanismo de transmiss˜ao da pol´ıtica monet´aria. O modelo VAR ´e utilizado para estimar e simular o efeito que a altera¸c˜ao da taxa de juro tem sobre a economia.
Duas conclus˜oes principais podem ser retiradas, primeiro, existe uma clara diferen¸ca nas respostas e no poder explicativo da taxa de juro nos modelos para Portugal antes e depois da ado¸c˜ao do Euro, uma vez que as repostas no modelo Pre-euro s˜ao de uma magnitude maior do que no modelo Pos-Euro. Por ´ultimo, existem semelhan¸cas entre a economia Portuguesa e Polaca, pois ambas demonstram respostas similares a choques na taxa de juro.
Palavras-chave: Bloco Ex´ogeno, Mecanismos de Transmiss˜ao Monet´aria, VAR C´odigo JEL: E53 e E58
I would first like to thank my thesis supervisors Professor Ryszard Kokoszczy´nsk of the Faculty of Economic Sciences at University of Warsaw and Professor Sergio Lagoa of ISCTE-IUL. The doors to Prof. Kokoszczy´nsk and Professor Lagoa offices were always open whenever I ran into a trouble spot or had a question about my research or writing. He consistently allowed this paper to be my own work but steered me in the right the direction whenever he thought I needed it.
Finally, I must express my very profound gratitude to my parents, grandparents and to my girlfriend for providing me with unfailing support and continuous en-couragement throughout my years of study and through the process of researching and writing this thesis. This accomplishment would not have been possible without them.
Thank you. Albino Muchacho
1 Introduction 12 2 Monetary Transmission Mechanism 14
2.1 Main Objectives of the Monetary Authorities . . . 14
2.1.1 Banco de Portugal . . . 14
2.1.2 Narodowy Bank Polski and European Central Bank . . . 14
2.2 Relevance of Monetary Transmission Mechanism . . . 15
2.2.1 Theoretical Foundation . . . 15
2.3 Problems in the Study of Monetary Transmission Mechanism . . . 16
2.4 Transmission Channels of Monetary Policy . . . 17
2.4.1 Interest Rate Channel . . . 18
2.4.2 Exchange Rate Channel . . . 18
2.4.3 Other Assets Prices Channel . . . 18
2.4.4 Credit Channel . . . 19
3 Literature Review 21 3.1 Short History of Monetary Policy Study With VAR Models . . . 22
3.2 Monetary Transmission Mechanism in Small Open Economies . . . . 22
3.3 VAR Studies for Portugal . . . 23
3.4 VAR Studies in Poland . . . 24
4 Empirical VAR Model 26 4.1 Changes Occurred in Portuguese Economy . . . 27
4.2 Endogenous Variables . . . 27
4.3 Exogenous Variables . . . 28
4.4 Stationarity Analysis . . . 29
4.4.1 Portugal . . . 29
4.4.2 Poland . . . 30
4.4.3 German Exogenous Variables . . . 31
4.5 Optimal Lag Length . . . 31
4.6 VAR Model . . . 32
4.7 Identification Scheme . . . 32
5 Empirical Results 34 5.1 Stability and Residuals Analysis . . . 34
5.2 Impulse Response Function Analysis . . . 34
5.2.1 Portugal Pre-Euro . . . 34
5.2.2 Portugal Post-Euro . . . 36
5.2.3 Poland . . . 39 v
5.3 Variance Decomposition . . . 43
6 Conclusions 47
References 49
A Descriptive Statistics and Graphs for Portuguese Variables 51 B Descriptive Statistics and Graphs for Polish Variables 54 C Descriptive Statistics and Graphs for German Exogenous Variables 57 D Stationary Tests for Portuguese Variables 59 E Stationary Tests for Polish Variables 74 F Stationary Tests for German Exogenous Variables 89 G Optimal Lag Length Tests 93 H Regression Output 95
I Stability Tests 98
J Correlation Table 104 K Variance Decomposition 105
3.1 Typical response to a positive monetary policy shock in SOE . . . 23
5.1 Accumulate response to a positive shock of one standard deviation point in interbank interest rate . . . 35
5.2 Accumulate response to a positive shock of one standard deviation point in interbank interest rate . . . 37
5.3 Accumulate response to a positive shock of one standard deviation point in interbank interest rate . . . 39
5.4 Accumulate response to a positive shock of one standard deviation point in interbank interest rate . . . 41
5.5 Variance decomposition with interbank effect . . . 44
A.1 Descriptive statistics for Portuguese variables . . . 51
A.2 Portuguese Unemployment rate . . . 51
A.3 Portuguese real GDP growth . . . 52
A.4 Portuguese inflation rate . . . 52
A.5 Portuguese 90 days interbank interest rate . . . 52
A.6 Portuguese growth rate of total credit given to non-financial corpo-rations . . . 53
A.7 Portuguese real effective exchange rate index based on unit labor cost 53 B.1 Descriptive statistics for Polish variables . . . 54
B.2 Polish Unemployment rate . . . 54
B.3 Polish real GDP growth . . . 55
B.4 Polish inflation rate . . . 55
B.5 Polish 90 days interbank interest rate . . . 55
B.6 Polish growth rate of total credit given to non-financial corporations . 56 B.7 Polish real effective exchange rate index based on unit labor cost . . . 56
C.1 Descriptive statistics for German exogenous variables . . . 57
C.2 German real GDP growth . . . 57
C.3 German inflation rate . . . 58
D.1 Augmented Dickey-Fuller test for Portuguese unemployment rate . . 59
D.2 Augmented Dickey-Fuller test for Portuguese real GDP growth . . . . 60
D.3 Augmented Dickey-Fuller test for Portuguese inflation rate . . . 60
D.4 Augmented Dickey-Fuller test for Portuguese 90 days interbank in-terest rate . . . 61
D.5 Augmented Dickey-Fuller test for Portuguese total credit given to non-financial corporations . . . 61
D.6 Augmented Dickey-Fuller test for Portuguese real effective exchange
rate . . . 62
D.7 First-difference augmented Dickey-Fuller test for Portuguese unem-ployment rate . . . 62
D.8 First-difference augmented Dickey-Fuller test for Portuguese real GDP growth . . . 63
D.9 First-difference augmented Dickey-Fuller test for Portuguese inflation rate . . . 63
D.10 First-difference augmented Dickey-Fuller test for Portuguese 90 days interbank interest rate . . . 64
D.11 First-difference augmented Dickey-Fuller test for Portuguese total credit given to non-financial corporations . . . 64
D.12 First-difference augmented Dickey-Fuller test for Portuguese real ef-fective exchange rate . . . 65
D.13 Phillips-Perron test for Portuguese unemployment rate . . . 65
D.14 Phillips-Perron test for Portuguese real GDP growth . . . 66
D.15 Phillips-Perron test for Portuguese inflation rate . . . 66
D.16 Phillips-Perron test for Portuguese 90 days interbank interest rate . . 67
D.17 Phillips-Perron test for Portuguese total credit given to non-financial corporations . . . 68
D.18 Phillips-Perron test for Portuguese real effective exchange rate . . . . 69
D.19 First-difference Phillips-Perron test for Portuguese unemployment rate 69 D.20 First-difference Phillips-Perron test for Portuguese real GDP growth . 70 D.21 First-difference Phillips-Perron test for Portuguese inflation rate . . . 70
D.22 First-difference Phillips-Perron test for Portuguese 90 days interbank interest rate . . . 71
D.23 First-difference Phillips-Perron test for Portuguese total credit given to non-financial corporations . . . 72
D.24 First-difference Phillips-Perron test for Portuguese real effective ex-change rate . . . 73
E.1 Augmented Dickey-Fuller test for Polish unemployment rate . . . 74
E.2 Augmented Dickey-Fuller test for Polish real GDP growth . . . 75
E.3 Augmented Dickey-Fuller test for Polish inflation rate . . . 75
E.4 Augmented Dickey-Fuller test for Polish 90 days interbank interest rate 76 E.5 Augmented Dickey-Fuller test for Polish total credit given to non-financial corporations . . . 76
E.6 Augmented Dickey-Fuller test for real Polish effective exchange rate . 77 E.7 First-difference augmented Dickey-Fuller test for Polish unemploy-ment rate . . . 77
E.8 First-difference augmented Dickey-Fuller test for Polish real GDP growth . . . 78
E.9 First-difference augmented Dickey-Fuller test for Polish inflation rate 78 E.10 First-difference augmented Dickey-Fuller test for Polish 90 days in-terbank interest rate . . . 79
E.11 First-difference augmented Dickey-Fuller test for Polish total credit given to non-financial corporations . . . 79
E.12 First-difference augmented Dickey-Fuller test for Polish real effective
exchange rate . . . 80
E.13 Phillips-Perron test for Polish unemployment rate . . . 80
E.14 Phillips-Perron test for Polish real GDP growth . . . 81
E.15 Phillips-Perron test for Polish inflation rate . . . 81
E.16 Phillips-Perron test for Polish 90 days interbank interest rate . . . 82
E.17 Phillips-Perron test for Polish total credit given to non-financial cor-porations . . . 83
E.18 Phillips-Perron test for Polish real effective exchange rate . . . 84
E.19 First-difference Phillips-Perron test for Polish unemployment rate . . 84
E.20 First-difference Phillips-Perron test for Polish real GDP growth . . . 85
E.21 First-difference Phillips-Perron test for Polish inflation rate . . . 85
E.22 First-difference Phillips-Perron test for Polish 90 days interbank in-terest rate . . . 86
E.23 First-difference Phillips-Perron test for Polish total credit given to non-financial corporations . . . 87
E.24 First-difference Phillips-Perron test for Polish real effective exchange rate . . . 88
F.1 Augmented Dickey-Fuller test for German real GDP growth . . . 89
F.2 Augmented Dickey-Fuller test for German inflation rate . . . 90
F.3 First-difference augmented Dickey-Fuller test for German inflation rate 90 F.4 Phillips-Perron test for German real GDP growth . . . 91
F.5 Phillips-Perron test for German inflation rate . . . 91
F.6 Phillips-Perron test for German inflation rate . . . 92
G.1 Optimal lag length for Portuguese model with German exogenous variables 1986Q1-1998Q4 . . . 93
G.2 Optimal lag length for Portuguese model with German exogenous variables 1999Q1-2017Q1 . . . 93
G.3 Optimal lag length for Polish model with German exogenous variables 1999Q1-2017Q1 . . . 94
H.1 Eviews output for Portuguese model with exogenous German block 1986Q1-1998Q4 . . . 95
H.2 Eviews output for Portuguese model with exogenous German block 1999Q1-2017Q1 . . . 96
H.3 Eviews output for Polish model with exogenous German block 1999Q1-2017Q1 . . . 97
I.1 Unit circle stability figure for Portuguese model with exogenous Ger-man block 1986Q1-1998Q4 . . . 98
I.2 Unit circle stability figure for Portuguese model with exogenous Ger-man block 1999Q1-2017Q1 . . . 98
I.3 Unit circle stability figure for Polish model with exogenous German block 1999Q1-2017Q1 . . . 99
I.4 Autocorrelation LM test for Portuguese model with exogenous Ger-man block 1986Q1-1998Q4 . . . 99
I.5 Autocorrelation LM test for Portuguese model with exogenous Ger-man block 1999Q1-2017Q1 . . . 100 I.6 Autocorrelation LM test for Polish model with exogenous German
block 1999Q1-2017Q1 . . . 100 I.7 Normal distribution errors test for Portuguese model with exogenous
German block 1986Q1-1998Q4 . . . 101 I.8 Normal distribution errors test for Portuguese model with exogenous
German block 1999Q1-2017Q1 . . . 102 I.9 Normal distribution errors test for Polish model with exogenous
Ger-man block 1999Q1-2017Q1 . . . 103 J.1 Correlation between Portuguese and EMU aggregated variables . . . 104 K.1 Variance decomposition of unemployment in the Portuguese Pre-Euro
model . . . 105 K.2 Variance decomposition of unemployment in the Portuguese
Post-Euro model . . . 106 K.3 Variance decomposition of unemployment in the Polish model . . . . 106 K.4 Variance decomposition of GDP in the Portuguese Pre-Euro model . 107 K.5 Variance decomposition of GDP in the Portuguese Post-Euro model . 107 K.6 Variance decomposition of GDP in the Polish model . . . 108 K.7 Variance decomposition of inflation in the Portuguese Pre-Euro model 108 K.8 Variance decomposition of inflation in the Portuguese Post-Euro model109 K.9 Variance decomposition of inflation in the Polish model . . . 109 K.10 Variance decomposition of interbank interest rate in the Portuguese
Pre-Euro model . . . 110 K.11 Variance decomposition of interbank interest rate in the Portuguese
Post-Euro model . . . 110 K.12 Variance decomposition of interbank interest rate in the Polish model 111 K.13 Variance decomposition of credit in the Portuguese Pre-Euro model . 111 K.14 Variance decomposition of credit in the Portuguese Post-Euro model . 112 K.15 Variance decomposition of credit in the Polish model . . . 112 K.16 Variance decomposition of exchange rate in the Portuguese Pre-Euro
model . . . 113 K.17 Variance decomposition of exchange rate in the Portuguese Post-Euro
model . . . 113 K.18 Variance decomposition of exchange rate in the Polish model . . . 114
4.1 Results of Portuguese stationarity tests . . . 29
4.2 First-difference results of Portuguese stationary tests . . . 30
4.3 Results of Polish stationary tests . . . 30
4.4 First-difference results of Polish stationary tests . . . 30
4.5 Results of German stationary tests . . . 31
4.6 First-difference results of German stationary tests . . . 31
4.7 Optimal lag length . . . 31
5.1 Effect of a positive monetary policy shock . . . 36
5.2 Effect of a positive monetary policy shock . . . 38
5.3 Effect of a positive monetary policy shock . . . 40
5.4 Elasticities to a unexpected monetary policy shock . . . 42
Introduction
Since the beginning of the Euro in 1999 that there is the question if a unique mone-tary policy is the best policy for the individual countries that belong to the European Monetary Union, especially for the more peripheric countries such as Portugal and Greece which are small peripheric economies with poor economic performances when compared to the core of Euro-Zone 1.
A major branch in monetary policy literature has been devoted to the study of this topic showing that despite the differences in terms of size and timing of the reaction to changes in interest rate across countries, in aggregate terms the monetary policy followed for the European Central Bank for the European Monetary Union is adequate 2.
Therefore, the objective of this study is not to evaluate if the monetary policy conducted by the European Central Bank is optimal for Portugal but rather if the relationship between six key macroeconomic variables 3 and the interest rates have changed or not as a consequence of the loss of monetary independence. This will be done by evaluating this changes and by how much these values shift.
Additionally, since some changes occurred in the Portuguese economy, apart from the loss of monetary policy independence, the analysis of Portugal was divided into two periods4. Three main changes happened, the decrease in the relative weight of the agricultural and industrial sectors and finally the increase in the relative weight of services in terms of total GDP.
Furthermore, the rest of the world, and in a closer look, the monetary trans-mission mechanisms have changed as a result of globalization and the increase of capital movements around the world, which caused an increase in the speed and size of the shocks that are spread worldwide. As a consequence of the 2008 finan-cial crises and then the 2010 European sovereign debt crisis lead to the adoption of unconventional monetary policy tools. As purchasing of European bonds by the European Central Bank in order to increase banks liquidity avoiding the liquidity trap. This dissertation, although not explicitly, also takes into consideration this unconventional monetary policy by adding credit to the model.
To take all of this into account, an exogenous block was introduced in the vector autoregressive model to capture changes occurred worldwide. In addition, Poland
1Germany, France and Italy
2(Ehrmann, Gambacorta, Martinez-Pag´es, Sevestre, & Worms, 2003) 3Unemployment, GDP, Inflation, Interest Rate, Credit and Exchange Rate 4Pre-Euro period (1986Q1-1998Q4) and Post-Euro period (1999Q1-2017Q1)
is used as a counter-factual 5 for the Portuguese Post-Euro period, since it shares
similar economic traces and has an autonomous and independent monetary policy. The structure of this dissertation is the following: in Chapter 2, is presented the objectives of three monetary authorities 6 and the main four transmission channels
of monetary policy.
In Chapter 3, a summarized literature review on one hand to stylize facts for small and open economies in another hand some studies for Portugal and Poland are revised. In Chapter 4, it is presented the reasons for choosing the variables, a series of experiments to test the stationarity variables of the model are conducted, followed by a discussion of the identification scheme.
In Chapter 5, consists on the presentation of the empirical results of the analysis, this is comprised of the impulse responses and the variance decomposition. Finally, in Chapter 6, it is showcased the synthesized conclusions of the empirical results.
5Although some caution is needed when analysing the results since the economies are similar
but not equal
Monetary Transmission
Mechanism
The objective of this chapter is to explain how the monetary transmission mechanism (MTM) operates and influences the economy through their channels. Each channel will be described in detail to highlight its importance in the economic activity.
2.1
Main Objectives of the Monetary Authorities
2.1.1
Banco de Portugal
The main objective of the monetary policy of Banco de Portugal (BdP), during the period 1986 to 1999 was too decrease the inflation rate and obtain price sta-bility. With this objective in mind the exchange rate policy began to be gradually less accommodative regarding the foreign shocks which caused an appreciation of the Portuguese escudo and generated positive real interest rates. The end of the crawling-peg policy in 1985, lead to the adoption of a policy that ensures exchange rate stability as a means of achieving price stability. This commitment of BdP was reinforced when the Portuguese escudo entered in the European Exchange Rate Mechanism in 1992 1.
2.1.2
Narodowy Bank Polski and European Central Bank
Both Narodowy Bank Polski (NBP) polish central bank and European Central Bank (ECB) share a common main objective, price stability in the medium-run and sup-port sustainable economic growth with low levels of unemployment2. Since the ECB began to implement monetary policy in euro area as of 1999 it has the objective of achieving an inflation rate below, but close to 2% . In a similar manner, in 2003 NBP started to target an inflation rate of 2.5% with a permissible fluctuation band of +/- 1 percentage points 3.
1(Abreu, 2005) 2(Issing, 2001) 3(Trembi´nska, 2010)
2.2
Relevance of Monetary Transmission
Mecha-nism
It is assumed that both monetary authorities, ECB and NBP, will behave in a similar manner since they share a similar objective, price stability by targeting the inflation rate to a specific value and use the interest rate as a tool to reach this goal. Therefore, the analysis will focus on ECB. To achieve the inflation objective the ECB must monitor all relevant developments in the 19-member states in the Euro area and act according to these developments, changing or not the current monetary policy. It is then crucial for the monetary authority, in this case the ECB, to understand the economic mechanism that generate adjustments in prices and output due to changes in the interest rate made by it.
Nevertheless, it is important to outline that each of the 19-member states have economic and financial systems operating at different stages. Thus, when the ECB decides to apply a specific monetary policy, different countries will react in different ways. Therefore, the monetary transmission mechanism has become much more complex due to different responses to one same policy.
2.2.1
Theoretical Foundation
It will be presented in a simple example, based in (Clements, Kontolemis, & Levy, 2001) and (Dornbusch, Favero, & Giavazzi, 1998) which allows for a simple and theoretical reaction function similar to the one used by ECB. For understanding processes it is assumed in the model that there are only two countries or two large set of countries that share similar characteristics. It is also assumed that during the design of the monetary policy, the ECB, wants to minimize a loss function identical to:
L = πeu2 = [θπ1+ (1 − θ)π2]2 (2.1)
where πeu represent the inflation in the euro area, π1 and π2 represent
respec-tively the inflation in country 1 and in country 2. The weight of the aggregation of indicators for member states is given by θ which varies between 0 and 1. The policy tool used by ECB is the interest rate which is used in the interbank money market as a reference. It is presumed that the relation between this special interest rate and the inflation for each country is given by:
π1 = −γ1r + 1 π2 = −γ2r + 2 (2.2)
r is the interest rate controlled by ECB, 1 and 2 represent the specific shocks
of each country and γ1 and γ2 are the transmission mechanism associated with a
particular country.
Substituting (2.1) in (2.2) and solving ∂L∂r = 0 in order to r. Obtaining the rule that minimizes the loss function:
r = θ1+ (1 − θ)2 θγ1+ (1 − θ)γ2
(2.3) which implies that a response of the interest rate to a shock in the inflation in country 1 will be given by:
∂r ∂1
= θ
θγ1+ (1 − θ)γ2
(2.4) As it is possible to showcase, the response of the monetary policy of ECB, due to the shock 1 will depend on the relative weight that country 1 has in the total
of the Euro area (which will also determine the importance of this shock in the final aggregated shock). Furthermore, the response of the monetary policy will also depend on the impact that it has on the two countries.
If country 1 has a small economy with a strong transmission mechanism (high γ), then if only country 2 suffer an inflation shock, the effect on inflation in country 1 will become bigger as γ1 increase. As it is shown by the following equations:
r = (1 − θ)2 θγ1+ (1 − θ)γ2 ; ∂r ∂2 = (1 − θ) θγ1+ (1 − θ)γ2 ; π1 = − 1 θ 1−θ + γ2 γ1 2
With these simple theoretical derivations, it is possible to observe that economic and monetary unions (EMU) will have difficulties to respond in an optimal manner to a shock. Specially, in a small country because it is unpredictable what the outcome might be. This is due to the fact that a big shock in a small economy will have a small influence on the aggregate indicators of the EMU. Therefore, the change in the interest rate will be insufficient to accommodate the big shock that happened in the small country. So, the change in the interest rate made by the monetary authority of the EMU will be smaller than the change that would happen in a Pre-EMU scenario, where the monetary policy was independent.
It will also be presented the case of independent monetary policy, exemplifying for the country 1. For sake of simplicity in the analysis, the assumptions made in the EMU case are still valid, especially the one that states the objective of monetary authority and the relation between inflation and interest rate. Then the reaction function of the independent monetary authority and the change of the interest rate needed to accommodate the shock in inflation will be given respectively by:
r1 = 1 γ1 (2.5) ∂r1 ∂1 = 1 γ1 (2.6) With this simple theoretical model is possible to illustrate that if different coun-tries with a non-homogeneous economic structure share a common monetary policy, the smallest countries will be the ones that will suffer the most in terms of output and inflation, the reaction of output and inflation will be even stronger for those small countries if they also have a strong monetary policy transmission mechanism.
2.3
Problems in the Study of Monetary
Trans-mission Mechanism
In the previous sub-chapter, it was presented why it is relevant and important to study the MTM. However, as any other topic in economics and other sciences, there are some issues in the study of this topic.
Those difficulties arise mainly from two sources4, the so-called problem of
simul-taneity and the effects of the different channels in MTM.
The simultaneity problem arises from the difficulties in the identification of the true effects induced in the economy due to changes in the monetary policy. This happens because monetary authorities adjust their policy to better accommodate the shock in the economy5. Therefore, it is hard to claim that a specific macroeconomic
variable changed just because of the intervention of monetary authority, having into account that the monetary authority might have just been responding to that shock. The following example can illustrate this problem, in the economy exists an inflationary pressure, the monetary authority will then tend to increase the interest rate, to cool down the economy, reducing the inflationary pressure. However, the prices will continue to increase, due to rigidities in the economy. Furthermore, there will be a positive correlation between the increase in the interest rate and in prices. Nevertheless, this increase in prices are linked to the past tendency and not because of the increase in the interest rates. In this case, simultaneous problem arises because it is hard to separate the two effects, the one that increases prices due to the previous tendency and the one caused by the changes in the interest rate. A common approach to overpass this problem is to study the effects of an unex-pected change in monetary policy, that is, a monetary policy shock. Nevertheless, this approach has some problems as well. With some authors claiming that empiri-cal research shows that these unexpected shocks are not one of the main reasons for the changes in the prices and output. Although, the same authors state that despite the small contributions in the variation in prices and output it cannot be said that monetary policy is not relevant 6.
The second difficulty referred above is associated with the complexity of the transmission mechanism that influence the economy. Since there are multiple chan-nels by which the monetary policy is transmitted to a particular variable, the fun-damental concern is in the correct identification of the effects that each channel has on a particular macroeconomic variable.
2.4
Transmission Channels of Monetary Policy
The study of the transmission mechanism and its’ channels is crucial for the under-standing of how the transmission of monetary policy influences the economy during the different stages in the economic cycle. This study is fundamental for a correct planning and implementation of the monetary policy. The final effect of these poli-cies are difficult to forecast, these difficulties are due to the time lag that exists between the implementation of the policy and the effects, also the magnitude of effects tends to be long and variable 7.
There are four main channels of transmission of monetary policy identified in the literature8 namely, the interest rate channel , the exchange rate channel, the credit channel and finally the other assets prices channel .
4(Kuttner, Mosser, et al., 2002)
5(Leeper, Sims, Zha, Hall, & Bernanke, 1996)
6(Leeper et al., 1996) and (Christiano, Eichenbaum, & Evans, 1999) 7(Friedman, 1995)
2.4.1
Interest Rate Channel
The interest rate channel is the primary channel by which the monetary policy is transmitted to the economy, in particular in the Keynesian models. This channel assumes that there is some degree of rigidity in prices in the short-medium run. Therefore, the monetary authority can influence real interest rate by manipulating the nominal interest rates, since prices are sticky by assumption. Thus, the inflation will not react instantaneously.
The increase in the nominal interest rate, due to a contraction in monetary policy, will lead to an increase in the real interest rate. The increase in the interest rate leads to a decrease in consumption and increase in the consumers’ savings, since now the savings are rewarded with the higher interest rate, which reduces the domestic demand for goods and services.
The consumer effect in this channel contributes to a decrease in the output of the economy.
2.4.2
Exchange Rate Channel
This channel explains how changes in the interest rate affect the exchange rate and by this the trade balance.
The adjustment in the interest rate causes a change in the nominal exchange rate due to the relative variation in the rewards of the assets denominated in the national and foreign currency. The adjustment of the real exchange rate will be similar to the change in the nominal exchange rate at least in primary stages. This occurs because of the nominal rigidity in the national economy in the short run.
Therefore, an increase in the nominal interest rate by the monetary authority will lead to an increase in the relative rewards on assets denominated in the national currency, this will cause an inflow of capital to the home economy. However this inflow is in the national currency, so the demand for national currency in the inter-national money markets will increase which will lead to an increase in the nominal exchange rate which will be passed to the real exchange rate. The main consequence of this increase in the real exchange rate is to make the products of the national currency to become more expensive for consumers abroad, therefore, there will be a decrease in the exports of the country. At the same time, the products that are based on foreign currency become now cheaper, which will reduce inflation so, the imports of the country will increase. These two effects will cause a deterioration on the trade balance which will cause a decrease in output.
This channel is particularly important for countries that have a small and open economies, such as, Portugal in the pre-EMU and Poland. However, due to the relatively big size of the Euro-Zone economy, this channel should not play an im-portant role in the transmission of the monetary policy of ECB in theory since the Euro-Zone is a relatively closed economy.
2.4.3
Other Assets Prices Channel
This channel explains how the wealth of families and firms’ investment is influenced by the monetary policy. Therefore, there are two effects in this channel on one hand the change of wealth of families and on the other hand the change in firms’
investment, measured by the Tobin q affect 9.
A contraction in monetary policy will increase the interest rates, which will lead to a decrease in the prices of financial (e.g. stocks and bonds) and real (e.g. real estate) asset that economic agents hold. Therefore, the decrease in the prices of those assets will reduce the consumption and investment.
The mechanism of transmission behind the first sub-channel is the assumption that the wealth is an important determinant of consumer expenditure. Therefore, a reduction in the wealth will reduce the level of consumption, which will lead to a decrease in output. It is expected that adjustments in real assets prices have a stronger effect than in financial asset prices. This happens because in countries such as Portugal and Poland the stock market has a lower importance in the aggregate saving of consumers.
The investment theory of Tobin (q effect) suggests a transmission mechanism that influences the stock market. The ratio q is defined as the market value of the firm divided by the cost of renewing capital, there is a positive correlation between this measure and the investment expenditure of firms 10. So, when there is an
increase in the interest rates by the monetary authority, the cost of renewing capital will increase, which will lead to a decrease in q, therefore, there will be a decrease in investment. Additionally, this increase in the interest rate will decrease the number of profitable projects which will decrease investment even more.
2.4.4
Credit Channel
The credit channel highlights the crucial role that banks have in the economic and financial systems. The effect of this channel is inversely correlated with the stage of development of financial and stock markets since, if these markets are on the early stage of development this means that most of the economy is financed via banks. This is exactly what happens in the Portuguese and Polish economy where the stock and financial markets are not the principal sources for financing the economy when compared with countries such as USA and United Kindom11, due to the fact
that Portugal and Poland are countries that are bank dominated, due to historical reasons and the lower development of financial markets in these countries.
The existence of credit rationing is caused by the operation of this channel. Fur-thermore, moral hazard and adverse selection problems faced by the banks during a contraction of monetary policy will increase credit rationing, causing a even stronger credit rationing which will lead to a decrease in investment and consumption. There-fore, a decrease in output. Nevertheless, this effect is independent of the demand for loans which is caused by the interest rate channel 12.
According to (Bernanke & Blinder, 1988), the credit channel assumes that there are frictions in the credit market, which will lead to the existence of the external finance premium. For this authors, the direct effects of the monetary policy in the interest rates are aggravated by the endogenous response of the external finance premium.
9(Mishkin, 1995)
10(Bolton, Chen, & Wang, 2011) 11(Allen & Gale, 2000)
In the literature are identified two main sub-channels in this channel, the Narrow Credit Channel and the Broad Credit Channel.
In the Narrow Credit Channel, the changes in external finance premium are due to changes in the supply of credit.
The increase in the compulsory legal reserves held in the central bank will reduce the amount available for credit. This reduction in the banking credit will increase its’ cost. Therefore, the aggregate expenditure of consumers will decrease due to the increase in the cost of credit. This effect will be bigger as the alternative sources of financing the economy are smaller.
The Broad Credit Channel operates throughout the monetary policy that influ-ences the financial position of the economic agents. Changes in monetary policy will influence the value of the collateral, therefore, the amount of credit given to the eco-nomic agents. In this way, an increase in the interest rate will lead to a decrease in the value of the collateral, which increases the default risk, due to problems such as, adverse selection and moral hazard. So, the creditor will increase the risk premium and reduce the amount of credit which will have negative effects on consumption and investment.
Furthermore, this mechanism can also be view in the perspective of the con-sumer’s expenditure. In this case, due to the decrease in the value of asset held by the consumers and the subsequent drop in cash-flows, it is expected that consumers decrease the consumption of durable goods, which normally are bought with credit so, the amount of credit demanded decreases.
Literature Review
The use of vector autoregressive models (VAR) has become a popular tool to analyse the Monetary Policy Transmission Mechanism. This type of model was firstly intro-duced by (Sims, 1980), in the famous article Macroeconomics and Reality. Due to the characteristics of this model, namely the fact that it does not require a specific or well define economic structure and the fact that the model does not require that variables are exogenous since it allows for the current values of a specific variable being affected by its own lag values and lag values of other variables the VAR models have become particularly suitable for the study of MTM.
Although the VAR models are one of the mostly used tools to study MTM, there are other approaches to study monetary transmission mechanisms, namely, DSGE and more refined models from the VAR family1. However, these models require more
data and a well define economic structure. This definition of economic structure is important for DSGE models, where all the economic structure and interactions among economic agents must be modelled.
As mentioned in the previous chapter, the understanding of MTM from the central bank is fundamental for a proper management of monetary policy, with this objective in mind ECB create the Eurosystem Monetary Transmission Network (EMTN) to study at the national and aggregate level the MTM in the Euro-Zone. This organism based their research in VAR models 2. Furthermore, NBP publishes
a two-yearly report called Monetary policy transmission mechanism in Poland in these reports the study of monetary policy is made by using (S)VAR models3, the
same models used by EMTN.
Another interesting feature in VAR models is that by design the model is able to only captures the response of endogenous variables to an unexpected change in monetary policy. For this reason and due to the model simplicity, this will be the empirical model that will be used to estimate how does the economy reacts to changes in the interest rate, that is, the impulse response function (IRF) of the Portuguese and Polish economy.
1For example SVAR and FAVAR models
2See for example (Van Els, Locarno, Morgan, & Villetelle, 2001), (Mojon & Peersman, 2001)
and (Peersman & Smets, 2001)
3See for example (Kapu´sci´nski et al., 2016), (Kapu´sci´nski et al., 2014) and (Demchuk et al.,
3.1
Short History of Monetary Policy Study With
VAR Models
As previously stated, the study of monetary policy with VAR was introduced by (Sims, 1980) with the study of US MTM case. Other authors try to replicate this study for other countries. Nevertheless, some puzzles appeared, namely, price, liquidity, exchange rate, and output puzzles. To try to solve these puzzles more variables were included in the VAR models. A successful example of this was the case of (Duguay, 1994) where the commodity index was introduced, and the puzzling results disappeared. This approach was followed by some authors by introducing a commodity index in the VAR models, even though, the economic intuition for the use of this variable was not always present. Nowadays, most of the study of MTM focus on SOE, by exploring the role of Exchange Rate and Credit channels in the economy.
3.2
Monetary Transmission Mechanism in Small
Open Economies
This section will be devoted to present the most common techniques and findings in the research on MTM in SOE.
Despite the lack of agreement between researchers regarding which is the best identification scheme to identify the effects of monetary policy in the economy, there is a consensus that there should be an exogenous block in the VAR models if the country is small and open to capture the fact that, the country has a relatively small impact on the world economy but is influenced by the world conditions through the financial markets and international trade. In general, the variables used on this exogenous block are the US GDP, as a proxy for the world demand and supply. Another variable commonly used in this block is the US inflation rate as a proxy for the world inflation 4. Another possibility is the use of the same variables for a local
partner which the country is dependent via trade or financial markets.
As mentioned above, there is some discussion regarding which is the best iden-tification scheme. Nevertheless, there is in the literature a dominant approached developed by (Christiano et al., 1999) which assumes for example a endogenous vec-tor of the following type5 Y
t= [yt, pt, mt, it, et] where Ytis the vector of endogenous
variables, yt is the real GDP, pt is the inflation rate, mt is a monetary aggregate, it
policy interest rate and et is the exchange rate. With this recursive ordering, it is
assumed that interest rate shocks do not have a contemporaneous impact on GDP, inflation and the monetary aggregate. Furthermore, it is assumed that changes in the interest rate will have a contemporaneous effect on the exchange rate. That is, it is assumed that the monetary authority observes at the beginning of the period the variables GDP, inflation and monetary aggregate and choose the interest rate according to the behaviour of those variables.
When the relative importance of the different channels of transmission of mone-tary policy is considered some stylized fact are well known, namely, the fact that the
4See for example (Cushman & Zha, 1997) and (Beier & Storgaard, 2006) 5For example (Peersman & Smets, 2001)
exchange rate channel is one of the most important channels of transmission of mon-etary policy to a small economy, particularly if the country is a small open economy. The role of the credit channel is inversely linked with the stage of development of the financial markets in that country since credit is an alternative source for financing the economic activity. Therefore, when the financial markets are not well-developed, the banking sector will have a higher influence on the economy. On the other hand, if the financial markets are well developed the other assets price channel will be more important for the transmission of monetary policy to the economy, because of the wealth effect present in this specific channel. Furthermore, it is possible to claim that there is a negative relationship between the relative importance of the other assets price channel and credit channel.
Figure 3.1: Typical response to a positive monetary policy shock in SOE Regarding the impulse response functions in most cases, they have the shape and signals that it is expected from them. As it can be seen in figure 3.1 for some variables, namely, output, prices, interest rate and exchange rate. With a tightening of monetary policy research and economic intuition show that there will be a decrease in GDP, in the money demand, on the amount of credit given to economic agents, and on inflation. Also, consistent with the overshooting hypothesis of (Dornbusch, 1976) there will be an increase in the exchange rate. Furthermore, typical response from economic and financial variables reach the pick after 4-8 quarters after the change in the monetary policy.
3.3
VAR Studies for Portugal
The VAR studies in Portugal were relatively uncommon before the integration of Portugal in the European Monetary Union. However, with the creation of EMTN by the European Central Bank, Portugal was included in most of the research in MTM done by this organization.
A good example is a research done by (Cecchetti, McConnell, & Quiros, 1999) for the EMU. The aim of the paper was to try to infer the relative weight that national central banks have been placing on output and inflation in the formulation of their policies, as well as, the trade-off coefficient of inflation-output implicit on the economic structure of each country.
The authors choose seven of the eleven countries that were the founders of the EMU, namely, Belgium, France, Germany, Ireland, Italy, Portugal, Spain and three non-EMU countries to be able to make comparisons, Denmark, U.K., and Sweden.
The authors then built a structural VAR with four variables: output, inflation, interest rate and exchange rate. Then the authors compute and analyse the impulse response functions (IRF) of output and inflation to a shock on the interest rate. The conclusion was that the IRFs were not similar across countries. The authors also analysed the relative weight of inflation and output for the formulation of the monetary policies in each of the countries. The main findings were once more that the policy standards were not equal across countries, with countries such as Portu-gal and Spain to put more weight on the output variability and countries such as Germany, Belgium, and France more preoccupied with the variability of inflation.
The main conclusion of the paper was that the process of obtaining inflation stability will be more costly for some countries than others since the transmission of monetary policy affects economies differently seeing that it is stronger in some countries than in others.
Similar findings are reported in (Ehrmann et al., 2003), in the set of studies done by EMTN. One of the examples of this studies published in the book is the study by (Ehrmann et al., 2003) for the period from 1980 until 1998 using output inflation and interest rate as endogenous variables,which shows that in aggregate terms the response of the Euro area was in line with the literature. Nevertheless, on a country level, in small countries such as Portugal output and inflation reaction were much stronger than in countries such as Germany or France, one of the explanations that the authors gave is that this result arises from the fact that this Portugal is a small country with strong transmission mechanism of monetary policy.
In a recent paper (Sousa, 2014) the author wants to study the financial markets and how monetary policy can influence them. For that, he uses data from 1947Q1 until 2007Q4 with the normal three endogenous variables plus commodity price index and stock price index. One of the main conclusions is that an increase in the interest rate will lead to a strong increase in unemployment, the drop in commodity price is faster than the inflation but both of them decrease strongly. Additionally, the author also shows that the Portuguese money demand is characterized by low output and interest rate elasticities.
Another study by (Georgiadis, 2015) builds a general VAR with every country from EMU using data from 1999Q1 until 2009Q4 with the following variables in the endogenous block: output, inflation, interest rate and exchange rate. However, the monetary policy instrument, the interest rate, is modelled as a function of the aggregate Euro area output and inflation. With this model, the author is able to show that due to rigidities in the unemployment and labour market, Portuguese output and inflation is particularly sensitive to changes in the interest rate.
3.4
VAR Studies in Poland
As previously referred, the NBP publishes every two years a report in which it analysis the monetary transmission mechanism in Poland. The reports started in 2011 and the last available is from 2017. Although some variables are introduced and drop in each report, the core variables are always presented, namely, GDP, inflation, interest rate, and the exchange rate. Furthermore, in every report, the
data is collected from the first quarter of 1998 until the year of the report. Therefore, it is possible to follow the different changes in the variables to a monetary shock and claim that this reactions are due to the recent adjustments in the Polish economy.
In (Demchuk et al., 2012) the authors built a structural VAR with the follow-ing variables on the endogenous vector: Consumer price index, investment, con-sumption, GDP, 1-month interest rate, and nominal effective exchange rate. After analysing the results from the model, the authors claim that the responses of the key macroeconomic variables to monetary policy were consistent with the structural features of the Polish economy and similar to the responses of the Euro area in ag-gregate terms. Although the financial system is less developed than in the Euro area, Polish economy is characterized by a lower degree of rigidities and more frequent price adjustments.
The same methodology is used in (Kapu´sci´nski et al., 2014). Also, the data set is expanded until the first quarter of 2013. Nevertheless, the order of the endogenous variables is different, with the investment and consumption being dropped from the model and the exchange rate is now real instead of nominal. An interesting result from the comparison between this report and (Demchuk et al., 2012) is the fact that the exchange rate channel seems to have decreased its relative strength and effectiveness, the authors argue that this shift is due to changes in the production process, related to the growing share of production by international firms. Another interesting result is that the response of inflation to changes in the monetary policy was now much stronger than what was reported in the previous study.
In (Kapu´sci´nski et al., 2016) same methodology was used as in the previous reports. The dataset was extended until the first quarter of 2015. In this report, the variable amount of loans denominated in zloty is included on the endogenous vector. Similar conclusions regarding the decrease of the relative importance of the exchange rate channel are found in this report. The inflation response to interest rate seems to be stable since similar responses were found in the previous report. One interesting result that comes from this study is that the authors tried to estimate the effect of the credit channel to the Polish economy and found that this particular channel has become relatively important to the transmission of NBP monetary policy to the economy.
As mention in the introduction, the objective of this study and therefore the research question is whether the responses of the Portuguese domestic variables have changed to shocks in the monetary policy due to the transference of monetary policy decision from Banco de Portugal to the European Central Bank. This evaluation it will in two steps, in the first step, there will be made a comparison before and after Portugal joined the Euro. In the second step, the responses of Portuguese and Polish domestic variables to shocks in the monetary policy will be compared.
It is expected that the reactions of Portuguese domestic variables to shocks in monetary policy to be different before and after the adoption of Euro, since if the European Central Bank increases the interest rates during a Portuguese recession this increase will depresses even more the economy. On the other hand, if this increase in the interest rate is made when Portugal is near the full employment this change will have a small impact on the economy.
Empirical VAR Model
The goal of this dissertation is to evaluate if the reaction of Portuguese economy to shocks in the monetary policy has changed after Portugal joined the Euro-Zone. Furthermore, a step further is made and compare the responses of Portugal with a European country that has autonomous and independent monetary policy.
This European country must be in the European Union, but outside of the Euro-Zone, additionally, it must have autonomous and independent monetary policy, that is, fluctuant exchange rate regime with inflation targeting, a country with a diversified economy with the agriculture to be relatively important in the 80s and 90s and same major commercial partners, to be as similar to Portugal as possible. With this characteristics, three countries emerge Sweden, Poland and Romania. However due to GDP per capita, Sweden is drop because it is too hight when compared with Portugal, similarly, Romania is dropped because GDP per capita in Portugal is too high. Therefore, the country that will be used to compare with Portugal is Poland. The methodology used for measuring the response of key macroeconomic vari-ables to shocks in the monetary policy is based on VAR models, it was considered the more adequate methodology since they are widely used for the study of unex-pected monetary policy shocks, it allows for the endogeneity of the variables, it is simple to implement and does not required a well specified economic structured as well as big data samples.
The data for Portugal was collected from the first quarter of 1986 until the first quarter of 2017. However, to make it possible the comparison between the response of key macroeconomic variables to shocks in monetary policy before the adoption of Euro and after the adoption of Euro, the sample is divided into two different periods. Although the number of observations is not the same in the two periods for Portugal, it is not expected that differences in responses are caused by this difference in the number of observations.
From the first quarter of 1986 until the fourth quarter of 1998 which will be denominated “Pre-Euro” during this period Portugal had an independent monetary policy 1. Also during this period, the Portuguese economy was characterized by
high-interest rates, high level of inflation and high growth rate of real GDP (except for the recession of 1993). The other period, the “Post-Euro” period the monetary policy was not independent since it was controlled by ECB. The Post-Euro data is from the first quarter of 1999 until the first quarter 2017, in this period the Portuguese economy was characterized by low-interest rates, low level of inflation
and low growth rate of real GDP.
The data for Poland was collected from the first quarter of 1999 until the first quarter of 2017. Furthermore, to make the comparison possible between Portugal and Poland the same variables are used in both models.
The data was collected from Federal Reserve of Saint Louis-Federal Reserve Eco-nomic Data for all the endogenous and exogenous variables with the exception of exchange rate which was collected from International Monetary Fund and Organi-zation for Co-operation and Development.
4.1
Changes Occurred in Portuguese Economy
Apart from the loss in monetary policy independence, several other characteristics have changed in the Portuguese economy before and after joining the Euro.
One of these changes was the decrease in terms of contribution for the total GDP in the 80’s and 90’s from sectors such as agriculture and fisheries which represented around 5.58% of the GDP and industrial sectors which accounts for 18.41% of GDP and after almost two decades the shares decreased to 1.85% and 13.11% respectively. On the other hand, the sector of services has increased the contribution share of GDP from 35.6% to 42.96%.
Another change that occurred was the fact that before the adoption of the Euro Portuguese economy was characterised by high inflation and as a consequence of that high interest rates. On the contrary, after the adoption of the Euro, the inflation was stable in low values which allowed to decrease the interest rates. This decrease in interest rates had two major consequences in the behaviour of families. On one hand, it created an incentive to increase credit which was given with lower interest rates than before. On the other hand, the decrease in the interest rates implied a decrease in the rewards of saving which decreased the families’ savings.
Finally, another changed that occur was the harmonization of the supervision rules among countries bellowing to the European Monetary Union by which com-mercial banks were supervised since this responsibility was passed to the European Central Bank which supervises the entire Euro system in coordination with national central banks.
4.2
Endogenous Variables
This study tries to evaluate the effect of monetary policy in the economy through three of four channels presented in the literature, namely, the interest rate channel, the exchange rate channel, and the credit channel. For this reason, six variables are collected: the real growth rate of GDP, unemployment rate, inflation rate, interbank interest rate, total credit, and real effective exchange rate.
The variable GDP is the real growth rate of the gross domestic product, mea-sured as the percentage change from the same period one year ago, this variable is also seasonally adjusted. This variable has been chosen since it is common that monetary authorities react to changes in output. Additionally, it is one of the ob-jectives of ECB and NBP to contribute to a sustainable economic growth if the inflation target value is not at risk.
The variable Un is the total harmonized unemployment rate seasonally adjusted. This variable has been chosen to control for the business cycle and it is effected by the rigidities present in Portuguese and Polish economies on the unemployment and labour market.
The variable Inf is the percentage change of consumer price index for all goods in relation to the same period one year ago and it is seasonally adjusted. This variable has been chosen due to the fact that is the most important key macroeco-nomic variable that the monetary authority targets and reacts to influence its future behaviour.
The variable Interbank is the 90 days interbank interest rate adjusted for brakes. This variable has been chosen since it is the key instrument that the mone-tary authorities in most advanced economies use to influence the economy and the economic behaviour.
The variable Cred is the percentage change from the same period one year ago of the total amount of credit given to non-financial corporations adjusted for brakes as well as seasonally adjusted. This variable has been chosen due to the fact that the credit channel has an important role in the Portuguese and Polish economy. Since those economies are banking dominated.
Finally, the variable Exch 2 is the real effective exchange rate index, based on
unit labour cost. This variable has been chosen due to the fact that both economies are relatively small and open. Therefore, the reaction of key macroeconomic vari-ables will also depend on the exchange rate.
The descriptive statistics, as well as the graphs for Portuguese variables, are presented in Appendix A and for Polish variables in Appendix B.
4.3
Exogenous Variables
As been mentioned, Portugal and Poland can be considered small open economies. To capture the idea that Portugal and Poland will react to shocks, for example, in German economy but this economy will not react to shocks in the Portuguese and Polish economy, it will be introduced in the VAR model an exogenous block.
In this block, there are two economic variables, GDP which is the real GDP as a measure of demand for national exports and supply for national imports. The other variable is inflation rate which is a proxy for the future foreigner inflation rate.
Two obvious countries can be representative for the rest of the world for Por-tuguese and Polish economies, US and Germany. The use of both countries to represent the rest of the world can be easily justified. In the case of US as the biggest economy in the world, it can be seen as a good proxy for world economic environment. On the other hand, Germany is the biggest commercial partner for both imports and exports to Portugal and Poland.
Therefore, some perspective analysis was made for the Portuguese model with U.S. and Germany for the Pre-Euro period, the response of key macroeconomic variables to changes in monetary policy was not that different. The analysis was then made for the Post-Euro period, here the conclusions were different. The responses of the key variables were much better the case where the rest of the world is represented by German economy. Same conclusions came from the Polish model.
Consequently, the country that will be used in the exogenous block in the VAR model will be Germany. The summary statistics, as well as the graphs for these two variables, can be consulted in Appendix C.
4.4
Stationarity Analysis
Before the estimation of the VAR model, it is important to study if the variables are stationary since VAR assumes that series are stationary.
To evaluate if a series is stationary there are multiple tests available. Never-theless, there will be used two different stationary tests to ensure robust results, namely, the augmented Dickey-Fuller test (ADF) and the Phillips-Perron test (PP). These tests measure the degree of integration of a variable. Furthermore, both tests have the null hypothesis that the variable has a unit root, that is, is not stationary. It will be now analysed the output generated by the econometric software EViews for all the variables in the analysis for the Portuguese and Polish economy as well as the exogenous variables3. The lags associated with both ADF and PP tests were
automatically chosen by the software. Furthermore, it was chosen the level of 1% significance level as a threshold to establish if a series is stationary. Also, if any of the tests show that the variable is not stationary first-difference operator will be applied to that variable to make it stationary.
4.4.1
Portugal
The ADF and PP test (in Appendix D) shows that none of the variables for the Portuguese economy is stationary, as it is possible to observe in table 4.1 .
Table 4.1: Results of Portuguese stationarity tests Variable ADF (p-value) PP (p-value Un 0.4751 0.6552 GDP 0.1559 0.0882 Inf 0.5705 0.2601 Interbank 0.7127 0.4577 Cred 0.0919 0.1037 Exch 0.2242 0.2086
Since none of the variables seems to be stationary, first-difference operator is applied to all variable, once more, the output of Eviews can be consulted in Appendix D, which are summarized in table 4.2.
3Most of the variables are in percentage change for this reason the option with no trend was
Table 4.2: First-difference results of Portuguese stationary tests Variable ADF (p-value) PP (p-value
D Un 0.0037 0.0000 D GDP 0.0000 0.0000 D Inf 0.0000 0.0000 D Interbank 0.0000 0.0000 D Cred 0.0000 0.0000 D Exch 0.0000 0.0000
As it is possible to observe in the table 4.2 the null hypothesis can be rejected for all the variables, that is, now all the variables are stationary in first-differences.
4.4.2
Poland
As in the case of the Portuguese variables, none of the variables are stationary with 99% confidence in both ADF and PP tests. Table 4.3 summarizes the results present in Appendix E.
Table 4.3: Results of Polish stationary tests Variable ADF (p-value) PP (p-value Un 0.7902 0.8686 GDP 0.0149 0.0192 Inf 0.1462 0.2986 Interbank 0.4871 0.6031 Cred 0.0337 0.0376 Exch 0.0126 0.0395
However, since VAR models require that all the variables are stationary, the first-difference operator will be applied to all variables that have a unit root. The results are available in Appendix E, which are summarized in table 4.4.
Table 4.4: First-difference results of Polish stationary tests Variable ADF (p-value) PP (p-value D Un 0.0089 0.0022 D GDP 0.0001 0.0000 D Inf 0.0002 0.0002 D Interbank 0.0003 0.0003 D Cred 0.0000 0.0000 D Exch 0.0000 0.0000
As it is possible to observe in the table 4.4 the null hypothesis can be rejected for all the variables, that is, now all the variables are stationary.
4.4.3
German Exogenous Variables
The ADF and PP test (in Appendix F) shows that only the variable GDP Germ is stationary, as it is possible to observe in the table 4.5.
Table 4.5: Results of German stationary tests Variable ADF (p-value) PP (p-value GDP Germ* 0.0000 0.0032 Inf Germ 0.1975 0.0992
* Stationary with 99% confidence for both tests in levels.
It is then needed to apply the first-difference operator to the German inflation variable that is not stationary in levels. The results are available in Appendix F which are summarized in table 4.6.
Table 4.6: First-difference results of German stationary tests Variable ADF (p-value) PP (p-value GDP Germ* 0.0000 0.0032 D Inf Germ 0.0000 0.0000
* Stationary with 99% confidence for both tests in levels.
As it is possible to observe in the table 4.6 the null hypothesis can be rejected for all the variables, that is, now all the exogenous variables are stationary.
4.5
Optimal Lag Length
Another step that is required to build a VAR model is to establish which is the adequate number of lags in the VAR model. The definition of a proper leg length is fundamental due to several aspects. On one hand, the optimal number of lags will ensure that the residuals of the model are stationary. On the other hand, the number of lags is what allows the VAR model to capture the economic dynamics of the country.
To guarantee robust results five criteria will be used to select the optimal lag length, namely, Likelihood-Ratio (LR), Final Prediction Error (FPE), Akaike Infor-mation Criterium (AIC), Schwarz InforInfor-mation Criterium (SC), and Hannan-Quinn Information Criterium (HQ) 4.
The results are available in Appendix G and are summarized in table 4.7. Table 4.7: Optimal lag length
Model LR FPE AIC SC HQ Portuguese 1986Q1-1998Q4 4 4 4 0 0 Portuguese 1999Q1-2017Q1 4 4 4 0 0 Polish 1999Q1-2017Q1 4 4 4 0 1
As it is possible to observe not all the criteria give the same optimal lag. Never-theless, three out of five criteria point that the optimal lag length is four quarters. Additionally, as a robustness check, a VAR model was estimated with three and five lags, the results were conclusive. The VAR with three lags were not able to generate IRFs with enough dynamics. In opposition, VAR with five lags demonstrate IRFs with much more volatility. Therefore, it is possible to conclude that the optimal number of lags is four.
4.6
VAR Model
As mentioned before, the VAR model will be a VAR model with four lags, VAR(4) with six endogenous variables and two exogenous variables. This VAR(4) can be represented by equation 4.1:
Yt= A0+ A1Y t − 1 + . . . + A4Yt−4+ B0+ Xt+ . . . + B4Xt−4+ t (4.1)
Which can be rewritten in the reduced form given by equation 4.2. Yt = A0 + 4 X i=1 AiYt−i+ 4 X i=0 BiXt−i+ t (4.2) Where:
Yt is a vector (6 x 1) with the endogenous variables of interest;
A0 is a vector (6 x 1) with the constant terms;
Ai is a matrix (6 x 6) with the endogenous coefficients;
Bi is a matrix (2 x 2) with the exogenous coefficients;
Xt−i is a vector (2 x 1) with the exogenous variables of interest;
t is a vector (6 x 1) with random terms, with mean 0, constant variance,
iden-tically and independently distributed and not correlated.
4.7
Identification Scheme
As mention in the literature review, the identification scheme developed by (Christiano et al., 1999) is widely used in the literature. Therefore, the identification scheme that will be used is inspired by those authors. The vector Yt has the order
repre-sented by expression 4.3: Yt= D Unt D GDPt D Inft D Interbankt D Credt D Excht (4.3)
In this setting, it is assumed that firstly, there is no contemporaneous impact of monetary policy on the unemployment rate, real GDP growth and inflation rate.
That is, the monetary policy tool, interest rate, does not influence the current values of those three variables just its future values. However, there is a possibility of credit and exchange rate to react contemporaneously to changes in the monetary policy tool.
In other words, the monetary authority observes the variables unemployment rate, real GDP growth and inflation rate and chose an interest rate that minimizes a similar loss function as shown in 2.2.1 and the remaining variables will react contemporaneously to this change in monetary policy.
Two main critics that can be made to this model. The use of many endogenous variables and the particular order of the endogenous vector Yt.
One of the principal problems of using too many endogenous is the fact that IRFs are not statistically significant. However, even in the most reduced VAR which only uses three endogenous variables, namely, GDP, inflation and the interest rate the statistical significance of the responses of the key macroeconomic variables to monetary shocks are similar to the model that is proposed.
Regarding the identification scheme of the endogenous variables, alternative schemes were tested. One of this tested schemes was similar to the one that is presented with the exception of the order of credit and interest rate. In this tification scheme the growth rate of credit came before the interest rate, this iden-tification scheme is sometimes used in the literature to capture the idea that the central bank reacts to changes in the credit market.
However, the result of this identification scheme and the others tested were qualitatively similar. Which indicate that the results that will be presented in the next chapter are robust to changes in the identification scheme.
As been mentioned, a natural step to build a VAR to a small open economy is to include an exogenous block to capture the idea that the economy is affected by foreign shocks but home shocks are irrelevant for foreign economies.
The exogenous block, Xt−i, will have the ordering describe in expression 4.4.
Xt−i =
GDP Germt−i
D Inf Germt−i
(4.4) This will be the fully specified VAR model that will be used to estimate the effect of changes in monetary policy in six key macroeconomic variables.
Empirical Results
Before analysing the impulse response functions and the variance decomposition, it is necessary to ensure that the model is well designed.
The output of the three models from the statistical software Eviews can be consulted in Appendix H.
5.1
Stability and Residuals Analysis
To verified if the properties of VAR model were respected, a series of stability test were made (Appendix I), namely, the unit circle test to verify if the VAR model is explosive, the LM test to ensure that residuals are not autocorrelated and finally, the normality of the residuals was tested.
As it is possible to observe, in the tests in Appendix I all the three models have a non-explosive behaviour since all residuals are within the unit root circle. Therefore, the results from the three models can be trusted and these results will tend to converge for the true values.
The LM test supports the idea that four lags is the optimal number of lags since it shows that there is no autocorrelation in the residuals when the model has 4 lags. The normality test shows that the residuals are normally distributed with the exception of the residuals of the Portuguese Post-Euro model. However, the law of big numbers ensure that residuals are normally distributed.
So, it is possible to conclude that the results from the three VAR models do not suffer from any statistical bias.
5.2
Impulse Response Function Analysis
In this section, it will be made the analysis of the impulse response functions fo-cussing on the effect of the interest rate channel. Firstly, it will be described briefly the behaviour of the IRFs for the three models. Then, it will be made a comparison between the behaviour of IRFs in the three models.
5.2.1
Portugal Pre-Euro
This period was used for verifying the ability of the identification scheme and the model to produce impulse response functions in line with the literature in small
